Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for bidding on placement of advertisements within media content over a computer networking environment, the method comprising: at each of one or more data management platforms, aggregating advertising and non-advertising data from one or more data suppliers, the advertising and non-advertising data including user profile data and media content profile data, wherein the user profile data comprises user characteristics and the media content profile data comprises media content characteristics, generating data mappings between the user profile data and user identifiers (IDs), generating data mappings between the media content profile data and universal resource locators (URLs), and transmitting the user profile data, the media content profile data, and corresponding data mappings to a first demand side platform; at the first demand side platform, storing the user profile data, media content profile data, and corresponding data mappings in a profile database, receiving one or more model files including a first model file, wherein each model file specifies a model for determining an adjustment score for adjusting performance goal values based on a combination of one or more of: user profile data and media content profile data associated with one or more advertisement requests, and receiving, from a first advertiser, a first set of campaign specifications for a plurality of first advertisements, wherein the first set of campaign specifications specifies: at least one first advertisement, a budget, the performance goal value, and the first model file; at an ad exchange server, receiving a first advertisement request, from a publisher server, for placement of an advertisement on a first advertisement space that is within media content that has been requested by a first user for display at a first user device, wherein the first advertisement request includes a combination of one or more of: a first set of user profile data associated with the first user and a first set of media content profile data associated with the media content, and transmitting the first advertisement request to a plurality of demand side platforms including the first demand side platform; at the first demand side platform, for the first advertisement request, inputting a combination of one or more of the first set of user profile data and the first set of media content profile data into the first model file to determine a first adjustment score, for each of the plurality of first advertisements, determining a bid for placement of such first advertisement on the first advertisement space, wherein each bid is determined based on at least the first adjustment score and the performance goal value, selecting a first bid from the plurality of determined bids, the first bid corresponding to a selected advertisement of the plurality of first advertisements, and sending the first bid, along with a reference to the selected advertisement, to the ad exchange server, wherein the reference includes a location of an ad creative file for the selected advertisement on a host server; at the ad exchange server, determining the first bid as a winning bid for finalizing purchase of placement of the selected advertisement on the first advertisement space, and sending the winning bid and the reference to the selected advertisement to the publisher server; and at the publisher server, retrieving the ad creative file from the host server, and transmitting the ad creative file to the first user device such that the selected advertisement is rendered for display at the first user device along with the requested media content.
This invention relates to a system for programmatically bidding on digital advertisement placements within media content across a networked environment. The system addresses the challenge of optimizing ad placement decisions by leveraging user and media content data to dynamically adjust bid values based on performance goals. The method involves aggregating advertising and non-advertising data from multiple sources, including user profile data (e.g., demographics, behavior) and media content profile data (e.g., content type, context). This data is mapped to user identifiers and media content URLs, then transmitted to a demand-side platform (DSP). The DSP stores this data and receives model files that define how to adjust performance goals (e.g., click-through rates, conversions) based on user and content characteristics. Advertisers submit campaign specifications, including ad creatives, budgets, performance goals, and the selected model file. When a publisher requests an ad placement for a user, the ad exchange forwards the request to multiple DSPs, including the first DSP. The DSP inputs the user and content data into the model file to generate an adjustment score, which modifies the performance goal value. The DSP then calculates bids for each ad in the campaign, selects the highest bid, and sends it to the ad exchange. The winning bid is finalized, and the publisher retrieves and displays the selected ad alongside the requested media content. This approach enhances ad targeting and performance by dynamically adjusting bids based on real-time data.
2. The method of claim 1 , further comprising: for each model file of the one or more model files, determining if the model file passes a set of predefined criteria; if the model file passes the set of predefined criteria, converting the model file into a corresponding executable model; determining if the corresponding executable model passes one or more runtime constraints; and if the model passes the one or more runtime constraints, storing the corresponding executable model for attaching to one or more sets of campaign specifications from one or more advertisers.
This invention relates to a system for processing and validating model files used in advertising campaigns. The technology addresses the challenge of ensuring that model files, such as those used for predictive analytics or ad targeting, meet predefined criteria before being deployed in live advertising systems. The method involves receiving one or more model files from a model provider and evaluating each file against a set of predefined criteria. If a model file meets these criteria, it is converted into an executable model. The executable model is then tested to ensure it complies with one or more runtime constraints, such as performance, accuracy, or compatibility requirements. Only models that pass these runtime checks are stored in a repository, where they can later be attached to campaign specifications provided by advertisers. This ensures that only validated, high-quality models are used in live advertising campaigns, improving reliability and performance. The system automates the validation process, reducing manual oversight and potential errors while maintaining strict quality control.
3. The method of claim 2 , wherein the set of predefined criteria includes a requirement that each model file only use a predefined set of variables as input for its specified model.
This invention relates to a method for managing model files in a computational system, addressing the challenge of ensuring consistency and reliability in model usage by enforcing strict input variable constraints. The method involves validating model files against a predefined set of criteria, where one key criterion requires that each model file only accepts a predefined set of variables as input for its specified model. This ensures that models operate within a controlled parameter space, reducing errors from unauthorized or unexpected inputs. The predefined set of variables is likely defined in a configuration or reference file, which the method checks against during validation. The method may also include additional criteria, such as format compliance or syntax validation, to further enforce model integrity. By restricting input variables, the method prevents inconsistencies that could arise from using undefined or incompatible variables, thereby improving the robustness and reproducibility of model outputs. This approach is particularly useful in environments where multiple models are deployed, ensuring that all models adhere to the same input standards. The method may be implemented as part of a larger system for model deployment, monitoring, or validation, where adherence to predefined criteria is critical for maintaining system reliability.
4. The method of claim 1 , further comprising: at the first demand side platform, receiving a second set of campaign specifications for a plurality of second advertisements, wherein the second set of campaign specifications specify a constant performance goal value and a plurality of constraints, for the first advertisement request, filtering the second advertisements based on whether the first set of user profile data and media content profile data meet each of the plurality of constraints specified by the second set of campaign specifications, and for each of the filtered second advertisements, determining a bid for placement of such second advertisement on the first advertisement space based on the constant performance goal value specified by the second set of campaign specifications.
This invention relates to digital advertising systems, specifically methods for optimizing ad placement in demand-side platforms (DSPs). The problem addressed is efficiently selecting and bidding on advertisements that meet campaign constraints while achieving performance goals in real-time bidding environments. The method involves a DSP receiving campaign specifications for multiple advertisements, where each campaign defines a constant performance goal (e.g., click-through rate, conversion rate) and multiple constraints (e.g., user demographics, content categories). When an ad request is received, the system filters advertisements based on whether user profile data and media content profile data meet all specified constraints. For each qualifying advertisement, the system calculates a bid value for placement in the available ad space, using the constant performance goal as a key factor in the bidding decision. This ensures that only relevant ads are considered, and bids are optimized to meet predefined performance targets. The method improves ad targeting efficiency by dynamically applying constraints and performance-based bidding, reducing wasted impressions and improving campaign effectiveness. It is particularly useful in programmatic advertising where real-time decision-making is critical.
5. The method of claim 1 , wherein the first set of campaign specifications for the first advertisements is associated with a plurality of constraints; wherein the method further comprises filtering the first advertisements based on whether the first set of user profile data and media content profile data associated with such each first advertisement meet each of the constraints; and wherein bids are only determined for the filtered first advertisements.
This invention relates to digital advertising systems that optimize ad selection and bidding based on user and content profiles. The problem addressed is the inefficiency in ad targeting and bidding processes, where ads may be shown to irrelevant users or content, leading to wasted ad spend and poor campaign performance. The system processes a first set of advertisements, each associated with user profile data and media content profile data. These ads are evaluated against a predefined set of constraints, such as demographic criteria, content relevance, or user behavior patterns. Only ads that meet all constraints are considered for bidding. This filtering step ensures that only the most relevant ads are presented to potential bidders, improving ad targeting efficiency and campaign performance. Additionally, the system may generate a second set of advertisements based on the first set, where the second set is optimized for different targeting criteria or bidding strategies. The second set is also filtered against the same constraints, ensuring consistency in ad selection. Bids are then determined only for the filtered ads from both sets, further refining the bidding process to focus on high-potential opportunities. By applying these constraints and filtering steps, the system enhances ad relevance, reduces wasted impressions, and improves return on ad spend. The method ensures that only ads meeting specific criteria proceed to the bidding phase, optimizing both targeting and budget allocation.
6. The method of claim 1 , wherein the performance goal value of the first model file is a cost per action (CPA) value.
The invention relates to optimizing machine learning models for advertising or marketing applications, specifically focusing on cost-efficient performance metrics. The method involves training a first machine learning model to achieve a performance goal, where the performance goal is defined by a cost per action (CPA) value. This CPA value represents the cost incurred per desired user action, such as a click, conversion, or other engagement metric, and serves as a key optimization target for the model. The method further includes generating a first model file from the trained model, which can be deployed to predict outcomes or make decisions based on input data. The trained model is designed to minimize the CPA, ensuring that the advertising or marketing campaign remains cost-effective while maximizing the desired actions. The method may also involve comparing the performance of the first model against other models or benchmarks to validate its efficiency. This approach is particularly useful in digital advertising, where balancing cost and performance is critical for campaign success. The invention addresses the challenge of optimizing machine learning models for real-world applications where cost efficiency is a primary concern.
7. The method of claim 1 , wherein the performance goal value of the first model file is a cost per click (CPC) value.
A system and method for optimizing machine learning models in digital advertising focuses on improving performance metrics such as cost per click (CPC). The invention addresses the challenge of efficiently selecting and deploying machine learning models to maximize advertising campaign effectiveness while minimizing costs. The method involves evaluating multiple model files based on performance goal values, where the performance goal value is specifically a cost per click (CPC) value. The system compares these values to determine the optimal model for deployment, ensuring that the selected model achieves the best possible CPC. The process includes generating model files from training data, calculating performance metrics for each model, and selecting the model with the lowest CPC to enhance advertising efficiency. This approach automates the model selection process, reducing manual intervention and improving campaign performance. The invention is particularly useful in digital advertising platforms where minimizing costs while maximizing engagement is critical. By focusing on CPC as a key performance indicator, the system ensures that the most cost-effective model is deployed, leading to better return on investment for advertisers. The method integrates seamlessly with existing advertising systems, providing a scalable solution for optimizing machine learning models in real-time.
8. The method of claim 1 , wherein the performance goal value of the first model file is a cost per thousand impressions (CPM) value.
A system and method optimize digital advertising performance by dynamically adjusting model parameters based on real-time data. The technology addresses inefficiencies in traditional advertising models, which often rely on static parameters that fail to adapt to changing market conditions or user behavior. The invention improves upon prior approaches by continuously evaluating performance metrics and automatically updating model configurations to maximize advertising effectiveness. The method involves training a machine learning model to predict optimal advertising parameters, such as bid values or targeting criteria, based on historical and real-time data. The model generates a performance goal value, which in this case is a cost per thousand impressions (CPM) value, representing the cost efficiency of an advertising campaign. The system monitors actual performance metrics, such as impressions, clicks, or conversions, and compares them to the predicted performance goal. If discrepancies are detected, the model parameters are adjusted to improve alignment with the desired CPM value. This iterative process ensures that advertising campaigns remain cost-effective and competitive in dynamic digital advertising environments. The invention further includes mechanisms for validating model updates, ensuring that adjustments do not negatively impact overall campaign performance. By continuously optimizing CPM values, the system enhances return on investment for advertisers while maintaining high-quality ad placements. This approach is particularly useful in programmatic advertising, where real-time bidding and automated decision-making are critical for success.
9. The method of claim 1 , wherein the first model file is specified as sharable and attached to first advertisements corresponding to different sets of campaign specifications from different advertisers.
A system and method for managing and distributing sharable model files in digital advertising campaigns. The technology addresses the challenge of efficiently sharing and reusing machine learning models across multiple advertising campaigns from different advertisers, reducing redundancy and improving performance consistency. The method involves creating a first model file that is designated as sharable, allowing it to be attached to multiple advertisements. Each advertisement corresponds to a distinct set of campaign specifications from different advertisers. The model file may include a trained machine learning model, such as a predictive model for ad targeting, performance optimization, or audience segmentation. By marking the model as sharable, advertisers can leverage pre-trained models without needing to develop or train their own, while the system ensures that the model remains consistent across different campaigns. This approach enhances scalability, reduces computational costs, and improves the accuracy of ad targeting and performance predictions. The system may also include mechanisms for version control, access permissions, and performance monitoring to ensure the model remains effective and secure across multiple campaigns.
10. The method of claim 9 , wherein the first model file is sold to the first advertiser or a model builder, wherein the first advertiser and model builder are not involved with creating the first model file.
This invention relates to a system for generating and monetizing machine learning models, particularly in the context of advertising. The problem addressed is the lack of a streamlined process for creating, distributing, and monetizing machine learning models, especially for advertisers or third-party model builders who do not participate in the model's initial development. The method involves generating a first model file using a machine learning process, where the model is trained on a dataset to perform a specific task, such as predicting user behavior or optimizing ad placement. The first model file is then sold to a first advertiser or a model builder, neither of whom were involved in the model's creation. This allows advertisers or model builders to acquire pre-trained models without needing to develop them in-house, reducing costs and development time. The system may also include additional steps, such as validating the model's performance, customizing it for specific use cases, or integrating it into existing advertising platforms. The monetization aspect ensures that the creators of the model can profit from their work while providing a valuable tool to advertisers and model builders. This approach democratizes access to advanced machine learning models, enabling smaller advertisers or businesses to leverage AI-driven insights without extensive technical expertise.
11. The method of claim 1 , wherein the model specified by the first model file is configured to determine the first adjustment score based on user characteristics specified by the first set of user profile data including one or more of the following: age, gender, and location of the first user.
This invention relates to personalized content delivery systems that adjust content recommendations based on user characteristics. The problem addressed is the need for more accurate and relevant content suggestions by incorporating specific user profile data into recommendation models. The system uses a model defined by a first model file to generate an adjustment score for content recommendations, where the model is configured to analyze user characteristics such as age, gender, and location from a first set of user profile data. These characteristics influence the adjustment score, which modifies the relevance of recommended content. The system also includes a second model file that defines a second model for generating a second adjustment score based on a second set of user profile data, allowing for further personalization. The adjustment scores are combined with a base relevance score to produce a final relevance score, which determines the ranking or selection of content for the user. This approach enhances recommendation accuracy by dynamically weighting user-specific factors in the content delivery process. The invention improves upon prior systems by leveraging detailed user profile data to refine content recommendations in real time.
12. The method of claim 11 , wherein the location of the first user includes a state and direct marketing area (DMA).
A system and method for targeted advertising leverages user location data to deliver personalized marketing content. The technology addresses the challenge of efficiently reaching specific audiences by utilizing precise geographic information to tailor advertisements. The method involves determining the location of a first user, which includes both a state and a direct marketing area (DMA). This location data is used to identify relevant advertising content that aligns with the user's geographic region. The system then selects and displays the targeted advertisement to the user based on this location information. The method may also involve tracking user interactions with the advertisement to refine future targeting strategies. By incorporating both state-level and DMA-level data, the system ensures that advertisements are highly relevant to the user's specific location, improving engagement and conversion rates. The approach enhances traditional location-based advertising by adding granularity through DMA segmentation, allowing for more precise audience targeting. This method is particularly useful for businesses that operate in specific regions or rely on localized marketing campaigns. The system may also integrate with other user data, such as browsing history or purchase behavior, to further refine the targeting process. The overall goal is to optimize advertising efficiency by delivering the right message to the right user at the right time, based on their geographic location.
13. The method of claim 1 , wherein each bid for each of the plurality of first advertisements is further determined based on a probability of a user action with respect to an advertisement.
This invention relates to online advertising systems that optimize bid placement for advertisements based on user engagement probabilities. The core problem addressed is improving the efficiency of ad placement by dynamically adjusting bids according to the likelihood of user interactions, such as clicks or conversions, rather than relying solely on fixed bid amounts or static metrics. The method involves analyzing historical or real-time data to calculate the probability of a user taking a specific action (e.g., clicking, purchasing) when presented with an advertisement. These probabilities are then used to adjust the bid amounts for each advertisement in a competitive auction environment. Higher probabilities of user engagement result in higher bids, increasing the likelihood of the ad being displayed in more favorable positions. Conversely, lower probabilities lead to reduced bids, conserving advertising budget for more promising opportunities. The system may also incorporate additional factors, such as user demographics, browsing behavior, or contextual relevance, to refine the probability calculations. By dynamically adjusting bids based on predicted engagement, the method aims to maximize return on ad spend while improving the overall relevance of displayed advertisements to users. This approach is particularly useful in programmatic advertising, where automated systems manage bid placements across multiple ad exchanges in real time. The invention enhances ad performance by aligning bid strategies with actionable user behavior insights.
14. The method of claim 1 , wherein the model specified by the first model file is configured to determine the first adjustment score based on different combinations of a plurality of user characteristics associated with different advertisement requests.
This invention relates to targeted advertising systems that optimize ad delivery based on user characteristics. The core problem addressed is improving the relevance and effectiveness of advertisements by dynamically adjusting ad selection based on user-specific data. The system uses machine learning models to analyze user characteristics associated with different advertisement requests and generate adjustment scores that influence ad selection. The method involves a model, defined by a first model file, that evaluates multiple combinations of user characteristics (such as demographics, behavior, or preferences) linked to different ad requests. The model computes a first adjustment score for each combination, which is then used to modify ad selection criteria. This ensures that ads are tailored to specific user segments, enhancing engagement and conversion rates. The model may be trained on historical data to refine its scoring mechanism over time. The system may also incorporate additional models or data sources to further refine ad targeting. For example, a second model file could define another model that adjusts ad selection based on contextual factors like time of day or device type. The combined insights from these models allow for highly personalized ad delivery, improving both user experience and advertiser ROI. The invention is particularly useful in digital advertising platforms where real-time decision-making is critical.
15. The method of claim 14 , wherein the different combinations of a plurality of user characteristics include different combinations of user demographics and different other contextual factors that affect user behavior differently.
This invention relates to a method for analyzing user behavior by evaluating different combinations of user characteristics, including demographics and contextual factors, to determine how these combinations influence behavior. The method involves collecting data on user demographics, such as age, gender, location, and occupation, as well as contextual factors like time of day, device type, and environmental conditions. By systematically testing various combinations of these characteristics, the method identifies patterns and correlations that explain variations in user behavior. The analysis may involve statistical modeling, machine learning, or other data-driven techniques to predict how different user segments will respond to specific stimuli, such as advertisements, product recommendations, or interface designs. The goal is to optimize user experiences by tailoring interactions based on these insights. The method can be applied in digital marketing, personalized content delivery, or user interface design to improve engagement and conversion rates. By considering multiple factors simultaneously, the approach provides a more nuanced understanding of behavior than traditional single-factor analyses.
16. The method of claim 1 , wherein the model specified by the first model file is configured to determine the first adjustment score based on different combinations of one or more of the following: user characteristics, media content characteristics, and advertisement characteristics.
This invention relates to a method for adjusting media content or advertisements based on dynamic scoring. The method involves using a machine learning model to generate an adjustment score that influences how media content or advertisements are presented to users. The model is configured to analyze different combinations of user characteristics, media content characteristics, and advertisement characteristics to determine the adjustment score. User characteristics may include demographic information, browsing history, or preferences. Media content characteristics may include genre, length, or popularity. Advertisement characteristics may include target audience, budget, or performance metrics. The model processes these inputs to generate a score that can be used to modify the presentation of media content or advertisements, such as adjusting playback order, visibility, or frequency. The method aims to optimize user engagement and advertising effectiveness by dynamically tailoring content delivery based on real-time data. The model is trained on historical data to improve accuracy over time. The invention addresses the challenge of delivering personalized media and advertisements in a way that balances user experience with advertising goals.
17. The method of claim 1 , wherein the first model file is received at the first demand side platform from the first advertiser.
A system and method for digital advertising involves a demand side platform (DSP) that receives and processes model files from advertisers to optimize ad placement. The DSP retrieves a first model file from a first advertiser, where the model file contains data used to predict the performance of ads. The DSP then uses this model to evaluate and select ad placements based on predicted outcomes, such as click-through rates or conversions. The system may also compare multiple model files from different advertisers to determine the most effective ad placements. The DSP may further adjust bidding strategies in real-time based on the model predictions to maximize advertising efficiency. The method ensures that ad placements are dynamically optimized according to the advertiser's performance models, improving campaign effectiveness. The system may also validate the model files to ensure they meet technical requirements before processing. This approach enhances ad targeting by leveraging predictive analytics, reducing wasted ad spend, and increasing return on investment for advertisers. The DSP may also integrate with other advertising platforms to expand reach and improve ad performance across multiple channels.
18. The method of claim 1 , wherein the user profile data is stored at the first demand side platform with an anonymized type of user identifier.
A system and method for managing user profile data in digital advertising involves storing user profile data at a demand side platform (DSP) using an anonymized user identifier. The anonymized identifier ensures user privacy while allowing the DSP to track and analyze user behavior across multiple advertising campaigns. The system integrates with a supply side platform (SSP) to facilitate real-time bidding (RTB) for ad placements, where the DSP submits bids based on the anonymized user profiles. The anonymized identifier prevents direct user identification but enables the DSP to maintain a consistent profile for targeted advertising. The method includes receiving user data from various sources, processing it to create anonymized profiles, and storing these profiles at the DSP. The system also supports bid requests and responses, where the DSP evaluates user profiles to determine optimal ad placements. The anonymized identifier ensures compliance with privacy regulations while enabling effective ad targeting. The overall system improves ad relevance and efficiency without compromising user privacy.
19. A system for bidding on placement of advertisements within media content over a computer networking environment, the system comprising memory and one or more processors configured to perform the following operations: at each of one or more data management platforms, aggregating advertising and non-advertising data from one or more data suppliers, the advertising and non-advertising data including user profile data and media content profile data, generating data mappings between the user profile data and user identifiers (IDs), generating data mappings between the media content profile data and universal resource locators (URLs), and transmitting the user profile data, the media content profile data, and corresponding data mappings to a first demand side platform; at the first demand side platform, storing the user profile data, media content profile data, and corresponding data mappings in a profile database, receiving a model file specifying a model for determining an adjustment score for adjusting a performance goal value based on a combination of one or more of: user profile data and media content profile data associated with one or more advertisement requests, and receiving, from an advertiser, a first set of campaign specifications for a plurality of first advertisements, wherein the first set of campaign specifications specifies: at least one first advertisement, a budget, the performance goal value, and the model file; at an ad exchange server, receiving a first advertisement request, from a publisher server, for placement of an advertisement on a first advertisement space that is within media content that has been requested by a first user for display at a first user device, wherein the first advertisement request includes a combination of one or more of: a first set of user profile data associated with the first user and a first set of media content profile data associated with the media content, and transmitting the first advertisement request to a plurality of demand side platforms including the first demand side platform; at the first demand side platform, for the first advertisement request, inputting a combination of one or more of the first set of user profile data and the first set of media content profile data into the model file to determine the adjustment score, for each of the plurality of first advertisements, determining a bid for placement of such first advertisement on the first advertisement space, wherein each bid is determined based on at least the adjustment score and the performance goal value, selecting a first bid from the plurality of determined bids, the first bid corresponding to a selected advertisement of the plurality of first advertisements, and sending the first bid, along with a reference to the selected advertisement, to the ad exchange server, wherein the reference includes a location of an ad creative file for the selected advertisement on a host server; at the ad exchange server, determining the first bid as a winning bid for finalizing purchase of placement of the selected advertisement on the first advertisement space, and sending the winning bid and the reference to the selected advertisement to the publisher server; and at the publisher server, retrieving the ad creative file from the host server, and transmitting the ad creative file to the first user device such that the selected advertisement is rendered for display at the first user device along with the requested media content.
This system enables targeted advertising by aggregating and mapping user and media content data to optimize ad placement. The system operates across multiple platforms, including data management platforms, demand side platforms (DSPs), ad exchanges, and publisher servers. Data management platforms collect advertising and non-advertising data, such as user profiles and media content profiles, from various suppliers. This data is mapped to user identifiers and URLs, then transmitted to a DSP. The DSP stores the data and receives campaign specifications from advertisers, including ad creatives, budgets, performance goals, and a model file for adjusting performance metrics. When a publisher server requests an ad placement, the ad exchange forwards the request to multiple DSPs, including the first DSP. The DSP uses the model file to generate an adjustment score based on user and media content data, then calculates bids for each ad in the campaign. The highest bid is selected, and the corresponding ad is sent to the publisher server, which retrieves and displays the ad alongside the requested media content. This system improves ad targeting by dynamically adjusting bids based on real-time data and performance goals.
20. At least one non-transitory computer readable storage medium having computer program instructions stored thereon that are arranged to perform the following operations: at each of one or more data management platforms, aggregating advertising and non-advertising data from one or more data suppliers, the advertising and non-advertising data including user profile data and media content profile data, generating data mappings between the user profile data and user identifiers (IDs), generating data mappings between the media content profile data and universal resource locators (URLs), and transmitting the user profile data, the media content profile data, and corresponding data mappings to a first demand side platform; at the first demand side platform, storing the user profile data, media content profile data, and corresponding data mappings in a profile database, receiving a model file specifying a model for determining an adjustment score for adjusting a performance goal value based on a combination of one or more of: user profile data and media content profile data associated with one or more advertisement requests, and receiving, from an advertiser, a first set of campaign specifications for a plurality of first advertisements, wherein the first set of campaign specifications specifies: at least one first advertisement, a budget, the performance goal value, and the model file; at an ad exchange server, receiving a first advertisement request, from a publisher server, for placement of an advertisement on a first advertisement space that is within media content that has been requested by a first user for display at a first user device, wherein the first advertisement request includes a combination of one or more of: a first set of user profile data associated with the first user and a first set of media content profile data associated with the media content, and transmitting the first advertisement request to a plurality of demand side platforms including the first demand side platform; at the first demand side platform, for the first advertisement request, inputting a combination of one or more of the first set of user profile data and the first set of media content profile data into the model file to determine the adjustment score, for each of the plurality of first advertisements, determining a bid for placement of such first advertisement on the first advertisement space, wherein each bid is determined based on at least the adjustment score and the performance goal value, selecting a first bid from the plurality of determined bids, the first bid corresponding to a selected advertisement of the plurality of first advertisements, and sending the first bid, along with a reference to the selected advertisement, to the ad exchange server, wherein the reference includes a location of an ad creative file for the selected advertisement on a host server; at the ad exchange server, determining the first bid as a winning bid for finalizing purchase of placement of the selected advertisement on the first advertisement space, and sending the winning bid and the reference to the selected advertisement to the publisher server; and at the publisher server, retrieving the ad creative file from the host server, and transmitting the ad creative file to the first user device such that the selected advertisement is rendered for display at the first user device along with the requested media content.
This invention relates to digital advertising systems that optimize ad placement using machine learning models. The system aggregates advertising and non-advertising data from multiple suppliers, including user profile data and media content profile data. It generates mappings between user profiles and user identifiers, as well as between media content profiles and URLs. This data is transmitted to a demand side platform (DSP), which stores it in a profile database. The DSP also receives a model file defining a model for adjusting performance goal values based on user and media content data, along with campaign specifications from advertisers, including ad creatives, budgets, performance goals, and the model file. When an ad request is received from a publisher for an ad space within media content requested by a user, the request includes user profile data and media content profile data. The DSP inputs this data into the model to determine an adjustment score, which is used to calculate bids for each ad in the campaign. The highest bid is selected, and the corresponding ad is sent to the ad exchange, which finalizes the purchase. The publisher retrieves the ad creative file and displays it alongside the requested media content. This system enhances ad targeting and performance optimization by leveraging machine learning models to dynamically adjust bids based on real-time user and content data.
Unknown
September 22, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.